Metal–thiolate bonds are present for many classes of metalloprotein active sites and make major contributions to function. The protein centers involved in electron transfer include the blue copper site (to be discussed in the next section), the mixed valent binuclear CuA site, and the Fe(SR)4, (Fe2S2)SR4, (Fe4S4)(SR)4, etc., iron sulfur sites all have thiolate–metal bonds of cysteine residues coupling them into the protein matrix.1 The main heme enzyme involved in O2 activation is P450, which has an axial thiolate–Fe bond, while enzymes involved in superoxide reactivity include superoxide reductase having a nonheme Fe–thiolate bond and Ni superoxide dismutase having nickel thiolate bonds.2 There are also classes of enzymes involved in Lewis acid catalysis that have thiolate–Fe bonds including nitrile hydratase (FeIII also CoIII, with some thiolates oxidized) and deformylase (FeII) or a thiolate-Zn bond as in alcohol dehydrogenase.3 Other metalloenzymes with bridging thiolate ligation include the hydrogenases, a binuclear (Fe–Fe or Ni–Fe) cluster involved in the reduction of dihydrogen, sulfite reductase, an FeIII porphyrin bridged to a Fe4S4 cluster involved in reduction of sulfite, and CO dehydrogenase, which has a binuclear Ni cluster bridged to a Fe4S4 cluster (A-cluster) is involved in the assimilation of CO (Fig. 1).1, 4
The general feature of all these metalloproteins and enzymes is that they have unique spectral features (i.e., intense, low-energy absorption bands and unusual spin Hamiltonian parameters) that reflect highly covalent thiolate–metal bonding that can make major contributions to reactivity. The thiolate ligand has three valence 3p orbitals, one of which is greatly stabilized in energy due to the carbon–sulfur σ bond, and thus does not significantly contribute to the thiolate sulfur–metal bond (Fig. 2). The two remaining 3p orbitals, which are perpendicular to the SC bond, dominate the thiolate interaction with the metal center and split in energy as the C–S–M angle decreases from 180°. The 3p orbital out of the C–S–M plane is involved in π bonding, while the in-plane 3p orbital pseudo σ bonds to the metal ion (pseudo σ in the sense that when the C–S–M angle is greater than 90° its electron density is shifted off the S–M bond; C–S–M bond angles are generally in the range of 100–120°). The specific bonding interactions depend on the metal ion, its 3dn configuration and its ZEff (effective nuclear charge).
In this review, we will first briefly consider the bonding in the blue copper site, which has one d hole and a high ZEff; therefore, it is a particularly covalent site. We then extend these studies over the series of first row transition metal ions to probe trends in covalent, ionic, π and σ bonding. Finally, we will consider FeIII–Sthiolate bonding, where the increase in effective nuclear charge for the ferric center again leads to highly covalent bonding and the high-spin d5 configuration allows both σ and π contributions to this bonding. Here we focus on the fact that the protein can provide hydrogen bonds to the thiolate, which can change the covalency of this bond and affect reactivity.
Nature of the Thiolate–Cu Bond in the Blue Copper Active Site
The details of this bonding description have been described,5 and only a brief summary of key features is presented here. The crystal structure of this site was first determined by Hans Freeman in 1978 for plastocyanin.6 It has a distorted tetrahedral geometry with a short Sthiolate ligand in the x,y plane, which is determined by this sulfur and two histidine nitrogen ligands all with strong ligand–metal bonds. There is also a weak axial thioether S ligand in some blue copper sites with a long bond to the copper (∼2.8 Å in plastocyanin7). The Cu–S–C bond angle for the thiolate is 110°, and this dihedral plane is oriented approximately perpendicular to the x,y plane. A DFT calculation of the ground state of the blue copper site was first published in 19858 (Fig. 3A), and shows a d orbital (i.e., oriented in the xy plane) which is highly covalent and the covalency is highly anisotropic being delocalized into the S 3pπ orbital of the thiolate ligand. Modern DFT calculations give an equivalent description but vary in the metal 3d-thiolate covalency, defined here as the amount of thiolate character mixed into the antibonding metal 3d-based molecular orbitals.
A direct experimental probe of the covalency of the thiolate S–Cu bond is sulfur K-edge X-ray absorption spectroscopy (XAS).9 This involves a transition at 2469 eV (Fig. 3B) from the sulfur 1s orbital into the half-occupied HOMO (i.e., SOMO or β-spin LUMO, which closely reflects the total spin density). The 1s orbital is localized on the sulfur atom and the s → p transition is the electric dipole allowed; thus, the intensity of the sulfur 1s → β LUMO transition directly reflects the sulfur 3p character mixed into the d orbital due to covalent bonding. The intensity of this transition for the blue copper site in plastocyanin is very high: 2.5 times the intensity of a model complex with a normal CuII–thiolate bond with ∼15% covalency. [In this model complex, the Cu atom has a five-coordinate geometry, Cu–Sthiolate distance is 2.36 Å, and the Sthiolate pπ → Cu and Sthiolate pσ → Cu LMCT transitions are at higher energy and much lower intensity—23,500 cm−1 (ε = 430 M−1 cm−1) and 27,800 cm−1 (ε = 360 M−1 cm−1), respectively, than those in blue-copper sites. Both the XAS intensity analysis and the SCF Xα calculations, calibrated to reproduce the EPR spin Hamiltonian parameters, give ∼15% S 3p character in the LUMO. This result and the error estimates in the S character from the XAS edges are discussed in detail in ref.10.] The intensity of the blue copper sulfur K-pre-edge transition quantitates to 38 ± 3% Sthiolate character in the ground-state wave function.5, 10
The π nature of the ground state in Figure 3A is quite unusual, as CuII normally utilizes its 1/2 occupied d orbital to form σ bonds to donor ligands. This π interaction derived from analysis of the charge transfer (CT) absorption spectrum using low-temperature magnetic circular dichroism (LT MCD) spectroscopy. From Figure 3C, the dominant transition in the absorption spectrum of blue copper proteins is a band at ∼16,000 cm−1 (ε ∼5000 M−1 cm−1), which is a characteristic unique spectroscopic feature of the blue copper site. Correlation to the LT MCD spectrum (Fig. 3D) shows that lower energy weak absorption features are the most intense in the LT MCD spectrum, while the intense 16,000 cm−1 band has only limited LT MCD intensity. LT MCD intensity (known as a C-Term) derives from spin-orbit coupling and reflects the metal character (which dominates the spin-orbit interaction) in the excited states. Thus, the low-energy weak absorption bands that are intense in LT MCD can be assigned as d → d transitions, while the 16,000 cm−1 band is the lowest energy; hence, π CT transition, while a higher energy weak feature is the Sthiolate pseudo σ to CuII CT transition. The intense-π/weak-σ CT absorption pattern is inverted from the normal behavior observed for CT transitions in copper complexes. As absorption band intensity reflects the overlap of the donor and accept orbitals involved in the electronic transition, this requires that the d orbital be oriented so that it π interacts with the thiolate sulfur. This π bond is due to the short thiolate S–CuII bond length of 2.1 Å.7 This short Cu–S bond originates from the fact that the Cu–SMet is very weak, and there are only three strong donor ligands (one SCys and two NHis) oriented in the xy plane of the copper coordination sphere. The high covalency of this bond activates specific superexchange pathways through the protein for rapid directional electron transfer as discussed in ref.5.
Ionic and Covalent, σ and π Contributions to MII–Thiolate Bonds
A series of metal-varied model complexes [MIIL(SC6F5)] (where L = HB(3,5-iPr2pz) and MII = Mn, Fe, Co, Ni, Cu, and Zn), related to blue copper sites in proteins, has been synthesized and crystallographically characterized.11, 12 The metal atoms in these complexes also have a distorted tetrahedral coordination sphere, with one M–S bond, two equatorial M–N bonds, and an elongated axial M–N bond. The spectroscopic features of [MIIL(SC6F5)] are similar to those of the corresponding metal-substituted sites in the proteins.13–20 The complexes show interesting systematic changes in their metal–thiolate bond lengths, which follow the order: MnII > FeII > CoII > NiII > CuII < ZnII. This parallels the Irving–Williams series for the stability constants.21, 22
Spectroscopic Signatures of π and σ Metal–Thiolate Bond Covalency
A combination of absorption, MCD, and resonance Raman (rR) spectroscopies was used23 to distinguish the ligand field (LF) transitions from the charge transfer (CT) transitions. The latter provide insights into covalent interactions at the metal center. Figure 4 compares the electronic absorption spectra of the [MIIL(SC6F5)] complexes. The experimental electronic spectra of the series are well reproduced by time-dependent density functional theory (TD-DFT)24, 25 at the B3LYP/6-311+G* level (Fig. 5) and confirm the assignments of the bands.
In contrast to the ZnII complex, which has the d10 metal ion configuration and does not show absorption in the visible region, the absorption spectrum of the CuII complex (Figs. 4 and 5B) has an intense absorption band at ∼15,000 cm−1, due to the Sthiolate pπ → Cu d− (β-spin HOMO → LUMO) CT transition. As described earlier, this is also characteristic of the blue copper proteins having the highly covalent Sthiolate pπ–Cu interaction in the ground state.5, 26, 27 The absorption spectrum of the NiII complex (Figs. 4 and 5C) exhibits a dominant feature at ∼20,000 cm−1, which is due to the Sthiolate pπ → Ni d− LMCT transition. This transition is higher in energy (from 15,000 cm−1 to 20,000 cm−1) and lower in intensity (from 12,000 M−1 cm−1 to 8000 M−1 cm−1) compared to the corresponding transition in the CuII complex (Figs. 4 and 5B). In the higher energy region (>25,000 cm−1) there are the Sthiolate pσ → Ni d−, benzyl based intraligand CT transitions, the Sthiolate pσ → Ni dz2, and the (pyrazolyl) pπ → Ni d− transitions.
In the absorption spectrum of [CoIIL(SC6F5)] (Figs. 4 and 5D), The Sthiolate pπ → Co d− charge transfer transition is now at 24,000 cm−1 (ε ∼3000 M−1 cm−1); this transition is higher in energy and lower in intensity compared to the corresponding band in the NiII complex (20,000 cm−1 and ε ∼ 8000 M−1 cm−1) as observed experimentally. The interesting feature in the CoII complex is an additional intense (ε ∼ 11,000 M−1 cm−1) Sthiolate pσ → Co dxy charge transfer transition at 30,000 cm−1. In CoII, the dxy orbital is unoccupied and interacts with the Sthiolate pσ orbital, producing an intense thiolate-to-metal CT absorption band at 30,000 cm−1. This reverses the pattern of more intense Sthiolate pπ and less intense Sthiolate pσ → MII d− CT transitions observed in the Cu and Ni complexes, to less intense Sthiolate pπ → MII d− and more intense Sthiolate pσ → MII dxy CT transitions. This behavior explains the spectral differences in Figure 4, where for the Co complex in solution the absorption intensity at ∼30,000 cm−1 due to the Sthiolate pσ LMCT transition is more intense than the transition at ∼25,000 cm−1 due to the Sthiolate pπ LMCT transition.
The TD-DFT calculated spectrum for the FeII complex (Fig. 5E) shows low intensity ligand field transitions in the 5000–20,000 cm−1 region. The Sthiolate pσ → Fe dxy LMCT excitation appears as the intense absorption (ε ∼ 8000 M−1 cm−1) at 32,000 cm−1, compared to the less intense (ε ∼ 2200 M−1 cm−1) Sthiolate pπ → Fe d− CT excitation contributing to the absorption intensity at ∼ 28,000 cm−1. The Sthiolate pσ → Fe dxy LMCT transition is higher in energy (by 2000 cm−1) and lower intensity (by 3000 M−1 cm−1), and the Sthiolate pπ → Fe d− LMCT transition is also higher in energy (by 4000 cm−1) and lower in intensity (by 800 M−1 cm−1) relative to the corresponding excitations in the Co complex.
The [MnIIL(SC6F5)] complex does not have absorption intensity in the 5000–25,000 cm−1 region (Figs. 4 and 5F). This is because the LF transitions in high spin Mn(d5) complexes are both spin and Laporté forbidden.28 The absorption intensity at ∼35,000 cm−1 is due to Sthiolate pπ → Mn d− LMCT transition, and at 38,000 cm−1 to the Sthiolate pσ → Mn dxy LMCT transition.
Thus, in both the experimental and calculated spectra, the interesting trends observed are: (1) in going from Cu → Ni → Co, the Sthiolate pπ → M d− transitions shift to higher energy and decrease in intensity, and (2) a new intense Sthiolate pσ → M dxy band appears in going from Ni to Co, which shifts to higher energy and decreases in intensity in going from Co to Fe to Mn. To obtain insight into the variation in the observed spectral features arising due to the different metal–thiolate interactions in these complexes, trends in the metal–thiolate bonding were evaluated using DFT.
Molecular Orbital Description of Metal–Thiolate σ and π Bonding and Ligand-to-Metal Charge Donation
Figure 6 shows MO diagrams for the [MIIL(SC6F5)] complexes, with the metal 3d orbitals and the thiolate ligand orbitals labeled. These MOs are relevant for chemical bonding between the metal atom and the thiolate ligand and define the optical spectra. The extended charge decomposition analysis (ECDA),29, 30 which is an extension of the charge decomposition analysis,31 indicates that the metal–thiolate covalent interactions in these series are dominated by charge donation from the SC6F ligand to the ML+ fragment (Table 1). This charge donation has both σ- and π-orbital components, and the two highest occupied molecular orbitals of the thiolate ligand, HOMO (S pπ) and HOMO-1 (S pσ), are principal donor orbitals for the MS bond.
Table 1. Extended Charge Decomposition Analysis29, 30 of the Metal–thiolate Bonding in the [MIIL(SC6F5)] Complexes: Thiolate HOMO(π) and HOMO-1(σ) Contributions to the Unoccupied MOs of [MIIL(SC6F5)], Net Contributions (%) of Unoccupied Fragment Molecular Orbitals (UFOs) to the Occupied MO of [MIIL(SC6F5)], the Difference in Electronic Polarization (PL) between the ML+ and SC6F Fragments, and the Net Charge Donation from the SC6F Thiolate to the ML+ Fragment (B3LYP/TZVP Calculations).
The principal component of π donation(SC6F → ML+): the HOMO(SC6F) contribution to the unoccupied MOs of the complex.
The main component of σ donation(SC6F → ML+): the HOMO-1(SC6F) contribution to the unoccupied MOs of the complex.
The difference (orbital %) in electronic polarization (PL) between the ML+ and SC6F fragments for α- and β-spin orbitals, respectively.
The net charge donation from the SC6F thiolate to the ML+ fragment for α- and β-spin orbitals, respectively. The former is largely σ donation and the latter is both σ and π donation.
Due to strong overlap between metal 4s and Sthiolate pσ fragment orbitals, the σ component of the MS covalent bond is substantial (Tables 1 and 2) and its strength remains approximately the same over the series. This is reflected by the observation that the α-spin occupied molecular orbitals of each [MIIL(SC6F5)] complex contain 26–32% of the unoccupied molecular orbitals of the ML+ fragment (Table 1) or, alternatively, the natural population analysis (NPA)32 derived occupation of the metal 4s orbital is 0.14–0.23 electrons (Table 2). Taking electronic polarization of the ML+ fragment into account (Table 1),29 compositions indicate the net σ charge donation of ∼0.5 e− from SC6F to the ML+ fragment.
Table 2. Metal and Sulfur NPA32-Derived Charges and Metal 3d, 4s, and 4p Atomic Orbital NPA Populations (β-spin)a in the [MIIL(SC6F5)] (B3LYP/TZVP Calculations).
NPA populations of metal 3d, 4s, and 4p α-spin orbitals are 4.92–4.99, 0.18–0.23, 0.01–0.03 electrons, respectively.
In the MnII complex, all the β-spin 3d orbitals of the central atom are unoccupied.
The MnII 3d–Sthiolate orbital interactions are weak and contribute little to the covalent bonding relative the stronger MnII 4s,4p–Sthiolate orbital interaction. The Mn 3d orbital character of the thiolate-based HOMO(π) and HOMO-3(σ) (the thiolate character is 95 and 86%, respectively) is 3 and 5%, respectively, and reciprocally the Sthiolate contribution is 2% for the Mn d− (dπ) and 3% for dxy (dσ) orbitals (Fig. 6).
In the FeII complex with one more valence electron (which occupies the β-spin dyz−xz orbital) than the MnII complex, the β-HOMO-LUMO gap becomes smaller, 3.7 eV (in comparison to the 4.4 eV gap in the MnII complex), reflecting the greater Z for Fe relative to Mn. The increased Z shifts the energies of metal-based orbitals closer to the occupied thiolate ligand orbitals and increases the covalent mixing between the metal and the thiolate (the Fe 3d characters in the Sthiolate pπ and pσ orbitals are 5 and 11%, respectively; Fig. 6).
In [CoIIL(SC6F5)], the next metal d orbital (dyz+xz) becomes occupied. Because Z for Co is greater than that of Fe, the β-HOMO-LUMO gap again becomes smaller (3.2 eV) and the covalency of the MSthiolate bond is further increased (the Co 3d character in Sthiolate pπ is 9%, Fig. 6). The important contribution to σ covalency comes from HOMO-5, which is a bonding combination of the Sthiolate pσ orbital and Co dxy. This agrees with the spectroscopic data (Figs. 4 and 5), which show an intense the Sthiolate pσ → Co dxy CT transition.
In [NiIIL(SC6F5)], the dxy orbital is now occupied. This cancels the Sthiolate pσ–M dxy bonding component for the covalent MS bond, eliminates the Sthiolate pσ → M dxy CT band (Fig. 5), and, as will be discussed later, the MSthiolate bond order decreases. Because the dxy orbital is now below the Sthiolate pσ orbital (Fig. 6), the latter is destabilized by ∼1 eV. The greaterZ for Ni results in a further lowering of the energies of M 3d-based orbitals and favors more efficient covalent bonding with the Sthiolate pπ orbital. As a result, the LUMO (d−) has 8% S contribution (Fig. 6).
In the CuII complex, the dz2 orbital is populated and the Cu d− is the only unoccupied d orbital. As a result of the still larger Z, the β-HOMO-LUMO gap for [CuL(SC6F5)] is lowest (2.4 eV) maximizing the orbital interaction between the Sthiolate pπ and Cu d− (Fig. 6). These two fragment orbitals mix to form the bonding β-spin HOMO [49% of HOMO(SC6F) and 41% β-LUMO(Cu d− of ML+)] and the antibonding β-LUMO [39% of HOMO(SC6F) and 44% β-LUMO(ML+)]. Alternatively, the Sthiolate character of the β-LUMO is 27% and the Cu 3d character of the β-HOMO is 21%. Thus, the β-HOMO/LUMO compositions in the complex indicate the most covalent MSthiolate bond in the series and largest net charge donation from the thiolate to the metal (Table 1). This agrees with the spectroscopic data (Figs. 4 and 5), which show an intense the Sthiolate pπ → Cu d− CT transition. In addition, ECDA29 (Table 1) shows that the electronic polarization of the ML+ fragment by the thiolate ligand is the largest for the CuII complex, 16 orbital%, in agreement with the smallest HOMO-LUMO gap of the CuL+ fragment in the series. This polarization provides additional stabilization to the highly covalent CuII–thiolate bond.
In the ZnII complex, all the d orbitals of the metal ion are occupied, including the Zn d−–Sthiolate pπ orbital. Thus, the π contribution to covalency is cancelled, the covalent bonding interaction between ZnII 3d orbitals and the thiolate orbitals becomes very small, and the charge donation from the thiolate to the metal is the lowest in the series (Table 1). The remaining covalent interaction comes mostly from the ZnII 4s orbital.
Ionic and Covalent Contributions to the Metal–Thiolate Bonds
The B3LYP calculations correlate well with the spectroscopic data, and can be used evaluate the trends in MS bond lengths, energies, and force constants (Fig. 7), and correlate these to the nature of bonding between the metal ion and the thiolate ligand.23
Overall, the B3LYP calculations reproduce experimentally observed structural changes in the series (Fig. 7A). The nonmonotonic variations in the MS bond lengths mark important changes in the metal–thiolate bonding, depending on the electronic configuration of the metal ion. The ML+–SC6F binding energy was calculated as the interaction energy between the metal–pyrazolyl fragment and the thiolate ligand:
In the [MIIL(SC6F5)] series, it ranges (Fig. 7B) from −124.7 kcal mol−1 (Fe) to −132.7 kcal mol−1 (Zn), and its variation does not mirror the changes in the MS bond lengths.23 MS force constant is the lowest for the MnII complex (1.13 mDyne Å−1), increases from MnII to FeII to CoII, where it reaches a local maximum, then decreases for the NiII complex, and increases for the CuII complex (the maximum value, 1.43 mDyne Å−1), and then decreases again for ZnII. The force constant (Fig. 7C) shows greater variation (26%) than the bond energy (6%). Its variation is more consistent with the changes in the MS bond lengths. The lack of correlation between the bond lengths and energies, and the force constants implies that there are varying contributions to the metal–thiolate bonding in this series, and a more detailed examination of the metal–thiolate binding energies and their ionic and covalent components is warranted.
The ML+–SC6F binding energy, Eo, can be partitioned into several contributions. First, Eo is separated into two components Eprep and Eint:
Eprep is the preparation (deformation) energy23, 29 necessary to transform the ML+ and SC6F fragments from their equilibrium geometries and electronic ground states to the those in the complexes:
Over this series, Eprep(ML+) and Eprep(SC6F) was 5.3–10.3 kcal mol−1 and 1.0–1.2 kcal mol−1, respectively. Eint is the interaction energy between the ML+ and SC6F deformed fragments. This interaction energy (Fig. 7B) can be further divided into two major components that can be interpreted in a physically meaningful way:
Here, Ecov is the covalent or orbital interaction energy (including the exchange repulsion energy33–35) and Eionic is the electrostatic interaction energy. The latter is estimated as a sum of electrostatic interactions between atomic charges, qNPA, from the two molecular fragments, ML+ and SC6F. The ESP-derived36 atomic charges could be also used to evaluate the electrostatic contributions to bonding. For an example, the Merz–Singh–Kollman ESP-derived atomic charges of noninteracting ZnL+ and SC6F fragments were similar to the NPA charges and, as a result, the ionic interaction energy calculated from the ESP fragment charges (−102.6 kcal mol−1) is close to the NPA-derived value, −104.5 kcal mol−1. However, the ESP charges in the [Zn(L)SC6F5] complex were different from both the NPA values and charges in the noninteracting fragments. This underlines the problem of finding ESP charges in large molecules using the standard electrostatic potential fitting algorithms.
The charge distribution in this calculation corresponds to the one in the complex, as opposed to the electrostatic interaction energy from the energy decomposition analysis of Kitaura–Morokuma33, 34, 37 and Ziegler,38 which is calculated with undistorted charge distributions corresponding to those in the isolated fragments. (If the undistorted charge distributions were used in the analysis, the ionic contribution would show very little variation with the metal because the reference state in the energy decomposition is the noninteracting ML+ and SC6F fragments.)
Because the metal–pyrazolyl fragment and the thiolate ligand carry opposite charges, the electrostatic interaction is attractive (Eionic < 0) in the [MIIL(SC6F5)] complexes and related to the atomic charges of the metal and the sulfur of the thiolate ligand (Table 2). It can be seen that the absolute values of metal and sulfur charges are highest in the MnII and ZnII complexes and lowest in the CuII complex. Thus, it can be expected that the ionic contribution to the MS bond energy is lowest in the CuII complex and highest in the MnII and ZnII complexes. Indeed, as will be discussed later in this section, the ionic bonding plays an important role in determining the overall strength of the ML+–SC6F interaction.
The analysis of covalent contributions to the chemical bonding can be done using Mayer bond orders,39BAB, and its α- and β-spin orbital components, B and B.23
where P and Ps are the density and spin-density matrices, respectively (P = Pα + Pβ; Ps = Pα − Pβ), Pα and Pβ are α- and β-spin electron density matrices, and S is the overlap matrix.
From the bond order analysis, the metal-sulfur bonding is a dominant interaction between the ML+ fragment and the thiolate ligand (Fig. 8A). The MS bond order (and the total ML–SC6F5 bond order) in the [MIIL(SC6F5)] complexes follows the Mn < Fe < Co > Ni < Cu > Zn progression found in the MS force constants (Fig. 7C) and the charge transfer from the SC6F ligand to the ML+ fragment (Table 1).
To understand this trend and to calculate the σ- and π-bond contributions, symmetry-adapted orbitals were used in the bond-order analysis:23
where BAB (Γ) is the bond order contribution from orbitals with irreducible representation Γ.
The α-spin LUMO of the ML+ fragment (this orbital has ∼80% M 4s character) remains relatively unperturbed (in its energy and composition) by the nature of the central atom. As a consequence, the mixing between the α-spin LUMO of the ML+ fragment and the HOMO-1 of the SC6F fragment and the resulting σ bond contributions to the Mayer bond orders from the α-spin MOs, BαMS (σ), remain similar for all complexes in this series. Because all α-spin metal 3d-based orbitals are occupied and cannot contribute to covalent bonding with the thiolate donor, the π bond component from α-spin MOs, BαMS (π), is close to zero (Fig. 8A).
Thus, BαMS (σ) and BαMS (π) provide a reference for changes in β-spin MO σ and π contributions to the bond orders (Fig. 8A). These changes result from an additional small σ bond and larger π bond contributions to the covalent bonding due to M dxy–Sthiolate pσ and M d−–Sthiolate pπ orbital interactions, respectively. The increasing Z (from MnII to ZnII) brings the energies of the dxy and d− orbitals of the metal atom closer to the thiolate occupied orbitals and makes the M 3d-thiolate covalency stronger, given that the electron occupancy of the appropriate MOs allows the net contribution to be positive. This is the case for the σ component, BβMS (σ), in going from MnII to FeII to CoII, and for the π component, BβMS (π), when proceeding from MnII to CuII. These changes in σ- and π-components of the covalent bonding between the metal and the thiolate produce the observed net bond order trends with a local maximum at CoII and a global maximum at CuII.
Following this analysis of the covalent bonding in [MIIL(SC6F5)], it is possible to explain their observed spectral features (Figs. 4–5). The shift to higher energy and the decrease in intensity of the strong Sthiolate pπ → M d− (M = Cu, Ni, Co, and Fe) CT transition is due to reduced M 3d–thiolate covalency, supported by the decrease in the pre-edge intensity in the S K-edge XAS data23 and in MO calculations. An important feature is observed in the Co complex, the Sthiolate pσ → Co dxy charge transfer transition is now present and more intense relative to the Sthiolate pπ → Co d− charge transfer transition. This is consistent with its ground state electronic structure description, indicating the strong interaction between Co dxy and the Sthiolate pσ orbitals. Similar to the trend in the Sthiolate pπ → M d− CT transition, the Sthiolate pσ → M dxy CT transition shifts to higher energy and decreases in intensity in going from CoII to FeII to MnII.
If the metal–thiolate bonding in [MIIL(SC6F5)] were limited to only covalent interactions, the metal–thiolate bond orders would directly correlate with the metal–thiolate interaction energies and, would show maxima of the metal–thiolate bond energies at CoII and CuII (note the correlation between the MS bond orders and Eint – Eionic; Fig. 8). However, although the CoII complex shows the second largest metal–thiolate binding energy (Fig. 7B), the absolute maximum is observed for the ZnII complex, not for the CuII complex. This indicates that the ionic contribution, along with the covalent component, plays an important role in determining the overall strength of the metal–thiolate interaction (Fig. 8C): ∼44% ionic and 56% covalent in MnII, becoming more and more covalent (with maxima at CoII for σ-type bonding and at CuII for π type), and back to ∼50% ionic and 50% covalent in ZnII.
The higher effective nuclear charge of the metal atom in going from Mn to Cu favors covalent metal–thiolate bonding but it also causes a decrease in metal ionic radii. The latter would cause a gradual increase in the ionic component of the ML+–SC6F bond energy in going from MnII to ZnII if the MS covalency were unperturbed. However, this effect is opposed by the changes in the MS covalency, which result in charge donation from the thiolate ligand to the ML+ fragment (Table 1) and cause the metal charge to become less positive and the sulfur charge to become less negative (Table 1). As a result, the ionic component of the MS bond energy decreases in going from MnII to CuII (Fig. 8C).
The metal–thiolate force constant is very similar in [CuL(SC6F5)] and [ZnL(SC6F5)]. The Zn complex shows a longer MS bond distance than the Cu complex; however, the calculated metal–thiolate interaction energy in the Zn complex is 2 kcal mol−1 stronger than in for the Cu complex with the largest MS covalency. For the Zn complex, the M 3d–thiolate covalency is lost (all 3d orbitals are occupied) and the remaining M 4s–thiolate covalency contributes ∼50% to the metal–thiolate interaction energy. This large ionic contribution compensates for the lost MS covalency and results in the strongest metal–thiolate bond in this series (Fig. 7B). The lack of a correlation between bond strength and length (Fig. 7) for the Cu vs. Zn complex reflects the differences in distance dependence of the ionic vs. covalent contributions to bonding. In such a case, the general correlation of the bond energy, length, and the force constant does not hold.
Interestingly, the metal–thiolate binding energies in [MIIL(SC6F5)] do not show the trend expected for the Irving–Williams series, which indicate that, if the successive stability constants of complexes of divalent ions of the first transition series are plotted against the atomic number of the element, there is a monotonic increase to a maximum at Cu irrespective of the nature of the ligand.21, 22 The calculated ML+–SC6F binding energy varies from −126.8 kcal mol−1 (Mn) to −130.8 kcal mol−1 (Cu), to −132.7 kcal mol−1 (Zn). As we discussed above, this is a result of the compensating effect in the metal–thiolate complexes where the covalent contribution to bonding is comparable in magnitude with the ionic contribution to bonding. To reconcile this fact with the experimentally observed Irving–Williams series, we have also analyzed the metal–ligand bonding in a series of [MII(HB(pz)3)(F)] complexes where the fluoride models the fourth ligand being replaced by the thiolate.23 Our analysis indicates that the metal–ligand bonding in these complexes is dominated by the ionic contribution and the metal–fluoride binding energy (calculated at the B3LYP/6-311+G(3df) level) decreases from −176.4 kcal mol−1 (Mn) to −166.1 kcal mol−1 (Cu) and then increases to −176.6 kcal mol−1 (Zn). This variation in the binding energy is a result of the increasing metal–ligand covalency in going from Mn to Cu, which reduces the ionic interaction. However, because of the dominant ionic contribution to bonding in the [MIIL(F)] series, the compensating effect of the increasing covalent bonding contribution is not as large as in the [MIIL(SC6F5)] series. As a result, the difference between the relative formation energies of the metal–thiolate complexes and the metal–fluoride complexes
shows the Mn ≈ Fe > Co > Ni > Cu < Zn progression (Fig. 8), which is the Irving–Williams series. Moreover, stability constants of the metal–thiolate complexes calculated from the binding energy differences are consistent with the experimentally observed quantities.21, 22
Thus, the “softer” thiolate ligand can have comparable covalent and ionic contributions to bonding and these compensate to produce little change in binding energy over the series of metal ions (open squares in Fig. 9). For the “harder” ligands (F−, OH−, H2O, etc.), the ionic term dominates and their binding energies are affected by changes in covalency over the series. It is the competition between these behaviors that produces the Irving–Williams series (solid circles in Fig. 9) in stability constants.
Nature of Fe–Thiolate Bonds
S K-edge XAS has been used to study the bonding in a series of divalent first row transition metal tetra-thiolates, and the results obtained are consistent with the general trend of bonding discussed above. Although CuII–Sthiolate exhibits a low energy intense pre-edge feature, FeII–Sthiolate has its pre-edge transition at a high energy due to its lower ZEff, and this overlaps the rising edge transition (Fig. 10, doted lines).
The S K-edge XAS of the ferric thiolate complex [FeIII(SPh)4]− (Fig. 10) has well-resolved low-energy intense pre-edges at ∼2470 eV.41 Using an effective S4 site symmetry, its pre-edge was fit using four peaks corresponding one-electron transitions to the low symmetry (Td → S4) split e(π) 5A + 5B (in red in Fig. 10 inset) and t2(σ) 5B + 5E (in blue in Fig. 10 inset) d6 excited states. (The peak splittings were fixed using an energy diagram determined from other spectroscopic techniques.41 The splittings were reduced by 20% to account for the fact that the final state has a reduced d6 configuration.) The fitted intensity gives a total hole covalency of 170% (summed over the five unoccupied Fe3d orbitals), which corresponds to the contributions from the thiolate S π and pseudo-σ from the four thiolates (∼43% per FeS bond). Spectroscopically calibrated SCF-Xα-SW calculations gave 140% total covalency in reasonable agreement with the experiment.40 The maximum π covalency was estimated to be 30% of the total observed covalency, which is consistent with the weak π and strong σ charge transfer transitions observed in the absorption spectrum for this complex.40 The redox active orbital in dz2, from DFT calculation, is metal based and has very little (6%) S character (as indicated by the S K-edge data and a weak π CT band in this complex).42 In contrast to the blue copper protein, which has extensive π delocalization enabling facile long range ET, the weak covalency of the redox active orbital in the rubredoxin site limits the electronic coupling of the metal into the protein and the ET process is localized in the vicinity of the active site as discussed in ref.42.
Effects of the Protein Environment on FeIII–Thiolate Bonding
The reduction potential of the iron–tetrathiolate site in rubredoxin is ∼1 V more positive than structurally similar model complexes. Hence, it is important to understand the contributions to this shift in redox potential of a protein active site that is key to its reactivity. Experimental and computational results suggest that H-bonds, protein dielectric effects, solvent accessibility, and surrounding peptide dipoles can make significant contributions to the redox potentials.43 S K-edge XAS has proved to be a powerful probe of H-bonding to sulfur ligands as this method directly probes the covalency of the FeS bond.
The S K-edge XAS spectra of rubredoxins from three different organisms, Clostridium pasteurianum (Cp), Pyrococcus furiosus (Pf), and a mutant having half the sequence from each of the above two proteins (Cp|Pf), are shown in Figure 11, along with data for the model complex [Fe(S2-o-xyl)2]−.41 The proteins display an intense pre-edge feature at approximately the same energy as the model complex. However, the intensity of this transition in the proteins is lower than that of the model complex, and this decrease in intensity varies somewhat in magnitude for the different proteins. This reduction in pre-edge intensity implies a reduction in FeSthiolate bond covalency, which was quantified by fits to the experimental spectra. The total Fe3d hole covalency for the four FeIII–Sthiolate bonds is 125–130% (∼32% S3p per bond) in the proteins compared to 170% in the model. The protein active sites have 6 backbone NH---Scys H-bonds which, from the pre-edge intensity, reduces the covalency of these sites relative to the model complexes.41
The amount of charge donation from the ligands to the metal, that is, covalency of the metal–ligand bond, can make a very significant contribution to the redox potential particularly in covalent systems like rubredoxin. Increased covalency preferentially stabilizes the higher oxidation state reducing the reduction potential. In the presence of H-bonds to the sulfur, the charge donation of the ligand to the metal should decrease and the redox potential becomes more positive. Although the S K-edge XAS data on the Rubredoxin proteins clearly demonstrate the effect of H-bonding in reducing the covalency in the protein active site, a more systematic evaluation of this effect was required.
H-bonding in a Series of FeIII–Sthiolate Heme Model Complexes
The effect of H-bonding on an Fe–Sthiolate bond was quantitatively estimated in series of high-spin FeIII pophyrin (P) model complexes FeP(SPh), FeP(SL1), and FeP(SL2) (where L1 = 2-trifluoroacetamido benzene thiol and L2 = 2,5-bis-trifluoroacetamido benzene thiol; Fig. 12), where the number of H-bonds was increased along the series (0 → 1 → 2) and a systematic variation in FeS bond lengths and reduction potentials was observed.44
The S K-edge XAS of the three model complexes [FeP(SPh), in bold], [FeP(SL1), 1H-bond in dots], and [FeP(SL2), 2 H-Bonds in dashed], are given in Figure 13.45 The data show that there is a decrease in pre-edge intensity along the series, and that the pre-edge peak maxima progressively shift to higher energy. (The energy shift was related to shifting of charge density away from the sulfur due to electron-withdrawing effect of the substituent on the phenyl ring. However, simulation showed that this-I effect of the substituent did not affect the FeS bond.) The energy and intensity of the pre-edge features are quantitatively estimated from fits to the experimental spectra and their second derivatives. The normalized intensity of the thiolate-based transitions is related to the total percent ligand character in the Fe3d antibonding manifold. This decreases from 1.30 for FeP(SPh) to 0.80 for FeP(SL2) corresponding to a decrease of FeS bond covalency from 49 to 31%. The t2-e orbital splitting in these complexes is not large enough to allow experimental resolution of the σ and π contributions to bonding.
DFT calculations were performed on the high-spin (S = 5/2) ground states of these heme complexes. The calculated geometries are in general agreement with the crystal structures and reproduce the FeS bond elongation on H-bonding as observed crystallographically (Table 3).44 The calculations can be correlated to the experimentally observed changes in pre-edge intensity. The MO diagram for the FeP(SPh) complex (Fig. 14) shows the dyz orbital has a π type interaction with the Sthiolate 3p orbital in the plane of the aromatic ring, and the dz2 orbital has a pseudo-σ-type interaction with the thiolate orbital out of the plane of the ring. In the crystal structure and the optimized geometry of the complexes, the thiolate binds such that the NH bonds are oriented directly toward the in-plane S 3p π orbital. The sum of Sthiolate 3p character in these β-spin unoccupied Fe 3d orbitals (Table 3) decreases from 42 to 30% from FeP(SPh) to FeP(SL2), paralleling the experimental results in Table 3. The calculations also indicate that the decrease in the thiolate contribution is solely in the π type orbital, consistent with the orientation of the H-bonds in these complexes.
Table 3. DFT (BP86/TZP in ADF 2004) Calculated Bond Lengths and Covalencies for the FeP(X) Complexes (Crystallographic Distance Are Given in Parenthesis).
Covalency (% S3p)
Total covalency (% S3p)
The energy of the H-bonding interaction was evaluated for both the aryl thiolate and a simplified alkyl thiolate (−SMe) with the same donor (H2O to model the H-bonding interaction from the ligand) and the results were the same.45 DFT calculations for FeP(SMe) and FeP(SMe) + 2H2O show that the FeS bond length increases on H-bonding by 0.02 Å in the oxidized form (Table 4 and Fig. 15). The covalency of the FeS bond decreases from 30 to 20% in the π orbital due to H-bonding (Table 4), similar to the effects found experimentally (Figure 13). The ΔE of H-bonding calculated for the ferric heme complex is about −5 kcal mol−1 for two H-bonds to the thiolate in the gas phase. This is in good agreement with previous estimates of the H-bond energy of sulfur donors.46
Table 4. DFT-Calculated FePSX Energies, MPA Populations, and Hirschfield Charges.
ΔEf (kcal mol−1) (solvent)
Energies obtained using the PCM method and CH2Cl2 as a solvent (G03 BP86/6-311+G*) are reported in parenthesis.
FeP(SPh) + 2H2O
FeP(SMe) + 2H2O
FeP(SPh) + 2H2O (red)
FeP(SMe) + H2O (red)
Energy of H-Bonding and Decomposition
This net change in energy (−5 kcal mol−1) is small considering the large change in FeS bonding interaction (decrease in π covalency by 33%; Table 3) involved in the process. The decrease of ligand–metal bond covalency should be reflected in the energy of metal–ligand bond more than in the total energy of the system. This is indicated in the energy decomposition given in Scheme 1, which shows that the bond energy of the H-bonded thiolate ligand and the FeIII-heme fragment is about −129 kcal mol−1, while that of the free thiolate is −151 kcal mol−1, 22 kcal/mol higher than the H-bonded ligand. Thus there is, in fact, a 22 kcal mol−1 decrease in BE of the FeS bond between the H-bonded and the non-H-bonded complexes that corresponds to the dramatic decrease in the covalent interaction observed from the S K pre-edge intensity (Fig. 13). However, the H-bonding energy of the thiolate with H2O is −27 kcal mol−1, which compensates for this difference in bonding energy and further stabilizes the system by the −5 kcal mol−1.
Similar DFT calculations performed on the one-electron reduced ferrous–heme complexes, show that there is an 0.05 Å increase in FeS bond length (Fig. 15) on H-bonding and the ΔE of H-bonding is about −12–15 kcal mol−1 for both alkyl and aryl thiolates. Thus, the H-bonding stabilization is ∼6 kcal mol−1 greater (Table 4) for the reduced relative to the oxidized complex for both alkyl and aryl thiolates after taking solvation into account (this stabilization is 10 kcal mol−1 in the gas phase). This energy difference contributes to the observed redox potential difference for these complexes. This H-bond energy difference predicts a 260 mV higher redox potential for the H-bonded couple. The experimentally observed difference in redox potential between the FeP(SPh) and FeP(SL2) couple is 330 mV. [The calculated FeII/III redox potentials for the FeP(SMe) and the FeP(SMe)+(2H2O) complexes are about −520 and −360 mV, respectively, in CH2Cl2. The measured redox potentials in the same solvent for FeP(SPh) and FeP(SL2) are −680 and −350 mV, respectively.]
In summary, the covalency of a FeSthiolate bond is about 42–49% without H-bonding and 32% with it, which is similar to that observed for the CuSthiolate bond in plastocyanin (38% with a single NH----Scys H-bond). The covalency of this bond is significantly reduced in the active sites of rubredoxins due to NH----S H-bonding from the backbone. This effect is quantitatively estimated in a series of high spin heme Fe–Sthiolate complexes where there is about a 30% decrease (49 to 32% in absolute numbers) of Fe–Sthiolate covalency for two H-bonds. DFT calculations reproduce these experimental trends and further show that the energy of H-bonding is small (−5 kcal mol−1 in the gas phase) compared to the large change in bonding observed in the S K-edge data. A bond energy decomposition scheme shows that the overall stabilization is a result of weakening in Fe–thiolate bond due to H-bonding coupled to the energy of H-bonding to the free ligand. The H-bonding stabilization increases by 10 kcal mol−1 (6 kcal mol−1 in solvent) in the FeII form due to higher charge density on the thiolate (less covalent FeII–Sthiolate bond), which increases the redox potential as observed experimentally.
Metal–thiolate bonds play a dominate role in bioinorganic chemistry. These can be highly covalent and π or σ (i.e., pseudo σ; Fig. 2) in nature depending on the effective nuclear charge of the metal ion and its dn configuration. The energy of the metal–thiolate bond is fairly insensitive to its ionic/covalent and π/σ nature as increasing covalency reduces the charge separation; hence, the ionic term and these contributions can compensate. Thus, trends observed in stability constants (i.e., the Irving–Williams series) then mostly reflect the dominantly ionic contribution to bonding of the innocent ligand being replaced by the thiolate. Due to their high effective nuclear charges of CuII and FeIII, cupric– and ferric–thiolate bonds are very covalent, with the former having strong π and the latter having more σ character. For the blue copper site, the high π covalency couples the metal ion into the protein for rapid directional long range electron transfer. For rubredoxins, because the redox active molecular orbital is π in nature, electron transfer tends to be more localized in the vicinity of the active site. Although the energy of hydrogen bonding of the protein environment to the thiolate ligands tends to be fairly small, H-bonding can greatly reduce the covalency of the thiolate–metal bond, which destabilizes the oxidized more than the reduced state and can significantly contribute to redox tuning by the protein environment.
S.I.G. is grateful to NSERC (Ottawa) for a postdoctoral fellowship. E.I.S. thanks his students and collaborators as cited in the references for their contributions to this research.